ALL SPEAKERS
AT A GLANCE
Role
Director of Engineering
Organisation
MoonPay
Category
Community
ABOUT THE SPEAKER
• Former Technical Product Manager at videro.com • Self-employed for 5 years with Photodesk-app.com, which reached the top 10 in the US App Store multiple times • Developing Hyperdeck.io for the past 2 years — a Markdown-based presentation app for professional pitches • Previously Team Lead Engineering at XING • Build tailoredpod.ai on the side • Currently Director of Engineering at MoonPay
TALK
The coming of local models
Most developers consider local models as a niche topic due to hardware requirements, performance constraints, and the simple fact that Claude or Codex are just plain better. This talk will show how this perspective becoming more outdated due to a variety of factors playing for a mix of cloud and local models.
First of all, modern models such as Qwen 3.6 27b or DeepSeek 4 Flash are close to capabilities of 3 to 6 months old cloud models. For some work this may not be sufficient, but for many tasks, this is good enough. Then, a motivated community of developers has continuously improved the quality & performance of local models through mechanisms such as DFlash, MTP, Steering or TurboQuant such that, for example the aforementioned Qwen went from 30 Tok/sec to 65 Tok/sec. Finally, not every company can pay the per/token costs of cloud providers of easily $1k+ per employee.
All this will lead more companies to use local models in parts of their stack
MORE VOICES
Other speakers at code.talks
See them live — and 100+ more.
One ticket, every track. Nov 4–5, 2026 at Kinopolis Hamburg.
Get your ticket





